Time Scale Algorithm: Definition of Ensemble Time and Possible Uses of the Kalman Filter

Abstract : The work reported in this paper takes place in a general theoretical overview concerning the generation of an ensemble time scale. Different algorithms can be designed to match the particular need of users and the available sets of clocks and time measurements. In all cases however, the statistical treatment of clock data requires at least: the definition of an average time scale, the spectrum of a procedure to optimize the contribution of each clock, the implementation of a filter on each clock frequency to provide a means of prediction. Here, the comparative study of two time scale algorithms, devised to satisfy different but related requirements, is presented. They are ALGOS(BIPM), producing the international reference TAI at the Bureau International des Pods et Mesures, and AT1(NIST) generating the real-time time scale AT1 at the National Institute of Standards and Technology. In each case, the time scale is a weighted average of clock readings, but the weight determination and the frequency prediction are different because they are adapted to different purposes. The possibility of using a mathematical tool, such are the Kalman filter together with the definition of the time scale as a weighted average, is also analyzed. Results obtained by simulation are presented.